INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

THE ACTIVITY of respiratory muscles, including the diaphragm (major inspiratory muscle), abdominal muscles (major expiratory muscles), and upper airway muscles (which act as valves to regulate airway resistance), is modulated during movement and changes in posture (3, 5, 8, 14, 18, 22). Recent studies in the cat (16, 17, 21, 28) have shown that part of the change in activity of respiratory muscles during alterations in body position is due to influences of the vestibular system. Electrical stimulation of the vestibular nerve produces responses in nerves innervating abdominal muscles, the diaphragm, intercostal muscles, tongue musculature, laryngeal muscles, and pharyngeal muscles (20, 21, 28). Selective natural stimulation of vestibular receptors, by rotation of the head (on a fixed body) in animals with denervations to remove neck, respiratory, cardiovascular, and facial somatosensory inputs that may be produced by the movement, produces modulation in the activity of nerves innervating abdominal muscles, the diaphragm, and muscles that move the tongue (16, 17). The effects of natural vestibular stimulation on other respiratory muscles have not been studied. Vestibular stimulation at small amplitudes (10-15° rotations) routinely produces responses in the abdominal muscles; in contrast, head rotations at amplitudes over 20° are sometimes required to affect neural outflow to tongue musculature and the diaphragm (16, 17). For all respiratory muscles studied thus far, the best direction of head rotation for producing an increase in activity is typically near nose-up pitch, although ipsilateral ear-down roll is also effective in activating some abdominal muscles in ~25% of animals (16, 17). The response characteristics of vestibular-respiratory responses suggest that they are due to stimulation of otolith organs and not semicircular canals (16, 17).

Several studies have considered the neural pathways that mediate vestibular influences on respiratory muscles. Vestibular-respiratory responses are abolished by chemical or mechanical lesions of regions of the medial and inferior vestibular nuclei caudal to Deiter's nucleus (16, 21, 28), suggesting that the medial and inferior vestibular nuclei are essential for producing these responses. Anatomic studies have shown that the medial and inferior vestibular nuclei project to regions of the lateral medullary reticular formation near the nucleus ambiguus and retrofacial nucleus containing neurons of the ventral respiratory group (VRG) (24). In contrast, the ventrolateral portion of nucleus solitarius containing dorsal respiratory group (DRG) neurons receives a paucity of projections from the medial and inferior vestibular nuclei (27). Electrophysiological studies confirmed that DRG inspiratory neurons receive little vestibular input (27) but that almost 50% of inspiratory bulbospinal VRG neurons (12) and over 80% of expiratory bulbospinal VRG neurons (20) respond to electrical stimulation of the vestibular nerve. However, chemical lesions that inactivated large portions of the VRG had little effect on vestibular-respiratory responses (30). In addition, transections of axons of VRG bulbospinal neurons, which are known to be the predominant source of expiratory signals to the spinal cord (2, 6, 7), reduce but do not abolish vestibular-abdominal responses (16, 20). These findings suggest that although VRG neurons receive vestibular inputs, other populations of neurons are also involved in relaying vestibular signals to spinal respiratory motoneurons.

As yet, the responses of VRG neurons to natural vestibular stimulation have not been characterized. It remains to be determined whether the responses of VRG neurons to natural vestibular stimulation will have the same spatial and temporal properties as vestibular-respiratory responses recorded from the diaphragm and abdominal muscles. One possibility is that VRG neurons have similar responses to natural vestibular stimulation as do spinal respiratory motoneurons but that additional populations of premotor respiratory neurons also have similar response properties, so that lesions of the VRG do not abolish responses of respiratory muscles to vestibular stimulation. Another possibility is that VRG neurons have substantially different responses to natural vestibular stimulation than do respiratory muscles and that more powerful vestibular signals from another group of neurons mask the vestibular inputs transmitted to spinal respiratory motoneurons by VRG neurons.

To determine the role of VRG neurons in producing vestibular-respiratory responses, we recorded activity from these neurons during whole body tilts in multiple vertical planes. In many animals, nonvestibular signals that might be produced by body movements were eliminated by transection of the IXth and Xth cranial nerves and the spinal cord at C₄. Similar nerve and spinal transections were previously shown to eliminate visceral inputs to the brain stem that were produced by whole body tilt (25, 26). In some animals without spinal cord transections, we also recorded activity from abdominal nerves during vertical tilts, so that vestibular-elicited changes in the activity of VRG neurons and spinal respiratory nerves could be directly compared. We additionally determined whether VRG neurons examined for vestibular inputs were bulbospinal by stimulating the C₂ spinal white matter to antidromically activate descending projections from the brain stem.

Some of these data have been reported in preliminary (abstract) form (23).

	METHODS

Top Abstract Introduction Methods Results Discussion References

All procedures used in this study conformed with the American Physiological Society's "Guiding Principles for the Care and Use of Animals" and were approved by the University of Pittsburgh's Animal Care and Use Committee.

General surgical procedures. Experiments were performed on 20 adult cats of either sex. Anesthesia was induced and maintained with 1-2% halothane (Fluothane, Ayerst Laboratories) vaporized in N₂O and O₂. Blood pressure was monitored from a femoral artery (with a Millar Mikro-Tip transducer), both femoral veins were cannulated to permit intravenous injections, and rectal temperature was maintained between 36 and 38°C with an infrared lamp and heating pad. If necessary, an intravenous infusion of lactated Ringer solution or metaraminol bitartrate (Aramine, Merck Sharpe & Dohme, 80 µg/ml) was used to keep blood pressure >100 mmHg. The animal was placed in a modified stereotaxic frame, with the head pitched down 30° to align the horizontal semicircular canals with the earth horizontal plane. As described below, this stereotaxic frame was mounted on a tilt table capable of simultaneous rotations in the roll and pitch planes. The animal's body was secured in place with the use of hip pins and a clamp placed on the dorsal process of the T₁ vertebra. A midcollicular decerebration was performed after ligation of the carotid arteries and aspiration of the portion of cerebral cortex overlying the brain stem. Approximately 1 cm of brain tissue rostral to the midcollicular transection was aspirated to ensure that decerebration was complete. A craniotomy was performed to expose the caudal cerebellum, and the caudalmost 2-3 mm of the cerebellar vermis were retracted or aspirated to allow access to the caudal brain stem. A laminectomy was performed to expose the C₂-C₄ spinal cord. Both hypoglossal nerves were identified, cut, and mounted in bipolar tunnel electrodes for recording of inspiratory activity (see below). In four animals, the L₁ nerve trunk innervating abdominal muscles (internal oblique, transverse abdominis, external oblique, and rectus abdominis) was dissected, cut, and mounted on a bipolar hook electrode.

Procedures to eliminate nonvestibular inputs that could be elicited by whole body tilt. In addition to activating vestibular receptors, whole body tilt may stimulate abdominal, cardiovascular, pulmonary, and other receptors. To exclude nonvestibular inputs that could be produced by body movement, the IXth and Xth cranial nerves were cut, and the spinal cord was transected at C₄ in 11 animals. We previously showed that these procedures are effective in removing visceral inputs to brain stem cardiovascular-regulatory neurons (25, 26), and so they presumably also eliminated tilt-related afferent signals other than those from the vestibular system to respiratory neurons. After the spinal cord transection, Aramine was always required to maintain mean blood pressure >100 mmHg.

Antidromic activation of neurons projecting to spinal cord. Three stainless steel floating electrodes, insulated to ~200 µm from the tip, were inserted into the lateral and ventrolateral white matter of C₂ on each side. We attempted to place each electrode at a slightly different laterality, so that the entire ventrolateral white matter would be stimulated. Monopolar square-wave current pulses 0.2 ms in duration and up to 1 mA in intensity were used for stimulation; the anode was attached to muscle adjacent to the stimulated cord. In every case, the antidromic nature of responses to spinal stimulation was confirmed using collision. In some animals, the locations of the electrode tips were marked by electrolytic lesions so that they could be reconstructed. The placement of electrodes in one animal is indicated in Fig. 1.

View larger version (14K): [in this window] [in a new window]   Fig. 1.   Camera lucida drawing of a transverse C2 spinal cord section showing locations of tips of stimulating electrodes (indicated by ×). Three electrodes were inserted on each side to maximize area of ventrolateral white matter that was stimulated.

Vertical vestibular stimulation. Vertical vestibular stimulation was produced by tilting the entire animal about the pitch (transverse) and roll (longitudinal) axes using a servo-controlled hydraulic tilt table (Neuro Kinetics, Pittsburgh, PA). The hydraulics of the tilt table were driven by sinusoidal stimuli delivered by a Cambridge Electronic Design (CED) 1401-plus data collection system interfaced with a Macintosh Quadra 800 computer. To characterize the vertical vestibular inputs to a neuron, we first determined the plane of tilt that produced maximal modulation of its firing rate (response vector orientation). Response vector orientation was determined using the "wobble" stimulus, a constant-amplitude tilt whose direction moves around the animal at constant speed (19). Clockwise wobble stimuli were generated by driving the pitch axis of the tilt table with a sine wave while simultaneously driving the roll axis with a cosine wave; during this stimulus, the animal's body, viewed from above, appeared to wobble, having in succession nose down, right ear down, nose up, and left ear down. When the signal to the pitch axis of the tilt table was inverted, the stimulus vector rotated in the counterclockwise direction. The direction of the response vector orientation lies midway between the maximal response directions to clockwise and counterclockwise wobble stimulation, because the phase differences between stimulus and response are reversed during the two directions of stimulation (19). Thus, by consideration of both responses, these phase differences can be accounted for. Wobble stimuli were delivered at frequencies ranging from 0.05 to 0.5 Hz (typically including 0.2 Hz) and at amplitudes up to 15°.

Recording of nerve activity. Activity was recorded from the hypoglossal nerves in all animals to allow monitoring of the respiratory cycle. Bursts of activity occur in the hypoglossal nerve in synchrony with inspiration, because the tongue must be protruded to maintain airway patency during the inspiratory phase of the respiratory cycle (2, 6, 7). In addition, we recorded from an L₁ abdominal nerve in four animals (with intact spinal cords), so that vestibular-elicited changes in the activity of brain stem respiratory neurons and a spinal respiratory outflow could be compared directly. Nerve activity was amplified by a factor of 10,000, filtered with a band pass of 10-10,000 Hz, full-wave rectified, and integrated (time constant of 100 ms for hypoglossal nerve and 1 ms for abdominal nerve). The signals were sampled at 500 Hz, stored, and displayed using the CED 1401-plus data collection system and Macintosh computer described above. Because of the long integrator time constant, inspiratory discharges recorded from the hypoglossal nerves appeared as an envelope of activity (see Figs. 2, 3, and 5 for examples) that could be easily compared with firing recorded from neurons.

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Examples of respiratory-related activity of neurons. In A-C, top traces show activity of a ventral respiratory group (VRG) neuron, and bottom traces show activity recorded from hypoglossal nerve. All traces are single, unaveraged sweeps. A: augmenting (Aug) activity recorded from an inspiratory (I; trace A1) and expiratory (E; trace A2) neuron. B: constant activity recorded from an I neuron (trace B1) and decrementing (Dec) activity recorded from an E neuron (trace B2). C: phase-spanning (PS) activity that begins during inspiration and continues into expiration (I-E, trace C1) and begins during expiration and continues into inspiration (E-I, trace C2).

View larger version (29K): [in this window] [in a new window]   Fig. 3.   Responses of an E-Dec cell to wobble stimulation. A: comparison of activity of neuron (top trace) with activity recorded from hypoglossal nerve (bottom trace). Both traces are single, unaveraged sweeps. B: histograms generated from 25 sweeps indicating response to clockwise (CW) and counterclockwise (CCW) wobble stimulation delivered at 0.1 Hz and at an amplitude of 15°. Responses to 2 directions of wobble stimulation were similar; in CW direction, signal-to-noise ratio was 0.6 and response gain was 0.4 spikes/s per degree, whereas in CCW direction signal-to-noise ratio was 1.0 and response gain was 0.6 spikes/s per degree. Response vector orientation calculated from responses to CW and CCW wobble stimulation was near nose-down pitch (97°). CED, contralateral ear-down tilt; IED, ipsilateral ear-down tilt; ND, nose-down tilt; NU, nose-up tilt.

Recording and analysis of unit activity. Electrode penetrations were made from 3 mm rostral to the obex to 4 mm caudal to the obex and from 2.5 to 4.5 mm lateral to the midline using epoxy-insulated tungsten microelectrodes with an impedance of 12 M (A-M Systems). Neural activity was amplified by a factor of 10,000, filtered with a band pass of 300-10,000 Hz, and led into a window discriminator for the delineation of spikes from single units. The output of the window discriminator was led into the CED 1401-plus data collection system and Macintosh computer; the sampling rate was 10,000 Hz. Electrolytic lesions were made in the vicinity of recording sites in each experiment so that recording locations could be reconstructed.

Histology. After the animals were killed, the brain stem and, in some cases, the upper cervical spinal cord were removed and fixed in formaldehyde solution. Sections (100 µm thick) were made in the transverse plane and stained with thionine. Locations of recorded neurons were reconstructed on standard sections with reference to placement of electrolytic lesions, relative locations of electrode tracks, and microelectrode depth.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

In total, 155 neurons with respiratory-related activity were tested for responses to vertical vestibular stimulation. Most of the units (102 or 66%) were recorded in animals with a transection of the spinal cord and IXth and Xth cranial nerves to eliminate nonvestibular inputs that might be produced by whole body tilt. Most (101) of the cells were tested for a projection to the spinal cord; 40 could be antidromically activated from the C₂ spinal electrodes (median threshold was 215 µA), whereas 61 could not be driven by a stimulus intensity of 1 mA.

Classification of respiratory-related activity of neurons. Neurons were classified as being respiratory-related if their spontaneous activity was synchronized with a phase of the respiratory cycle; the respiratory cycle was monitored by recording inspiratory activity from the hypoglossal nerve. Cells that fired only during bursts of activity in the hypoglossal nerve were classified as inspiratory (I); cells that fired between discharges in the hypoglossal nerve were classified as expiratory (E); and neurons that fired during a portion of both the inspiratory and expiratory phases were classified as phase spanning (PS). I and E neurons were subclassified as having augmenting, constant, or decrementing activity, as described in previous reviews (2, 6, 7). Examples of several different types of respiratory-related activity are illustrated in Fig. 2. Table 1 indicates the number of cells with different types of respiratory activity that were tested for a response to tilt. The sample was comprised predominantly of I (52%) and E (38%) neurons; PS neurons accounted for only 10% of the population. Among the E and I neurons, the augmenting discharge pattern was by far the most common. It was not possible to accurately subclassify the respiratory discharge pattern for some I and E neurons, particularly those with a slow firing rate.

Responses to sinusoidal tilt of neurons with respiratory-related activity. Typically, neurons with respiratory-related activity were first tested for responses to both clockwise and counterclockwise wobble stimuli delivered at 0.2 Hz and at an amplitude of 7.5°. If this stimulus was ineffective, we also routinely tested for responses to 0.1-Hz stimuli at amplitudes of 7.5-15° (maximal amplitude employed depended on mechanical stability of recording). As indicated in Table 1, only 25% of the I neurons, 19% of the E neurons, and 25% of the PS neurons responded to these stimuli. For the neurons whose activity was not modulated by vertical vestibular stimulation, approximately one-half were tested for responses to 15° stimuli at 0.1 Hz (21 of 48 E cells, 27 of 60 I cells, and 6 of 12 PS cells). The fraction of VRG neurons whose activity was modulated by tilt was similar in animals with transection of the spinal cord and IXth and Xth cranial nerves (27 of 102 cells) to that in animals without these transections (8 of 53 cells). The characteristics of responses to tilt were also similar in the two populations of neurons. It thus seems unlikely that visceral signals contributed appreciably to the responses of VRG neurons to tilt in those animals without transections of the spinal cord and IXth and Xth cranial nerves. The responses recorded from all VRG neurons studied were therefore pooled for subsequent analyses.

View larger version (22K): [in this window] [in a new window]   Fig. 4.   Plot of response vector orientations for all neurons whose activity was modulated by tilt. Neurons classified as receiving different types of vestibular inputs are plotted at different distances from center of graph. Approximate planes of vertical semicircular canals are midway between roll and pitch planes and are indicated by dashed lines.

View larger version (32K): [in this window] [in a new window]   Fig. 5.   Responses of an I-Aug neuron to tilts at multiple frequencies in plane of ipsilateral anterior and contralateral posterior semicircular canal. A: comparison of spontaneous activity of neuron (top trace) to activity recorded from hypoglossal nerve (bottom trace). B: responses of neuron to planar tilt at multiple frequencies; stimulus amplitudes employed were 7.5° at 0.2 and 0.5 Hz, 5° at 1 Hz, and 2.5° at 2 Hz. Histograms represent average of 29 sweeps at 0.2 Hz, 56 sweeps at 0.5 Hz, 100 sweeps at 1 Hz, and 251 sweeps at 2 Hz. However, relative amplitudes of responses were adjusted to account for the fact that a different number of sweeps was averaged in each trace. Arrows indicate when stimulus position was maximal in direction of anterior or posterior semicircular canal (i.e., when tilt amplitude was maximal). C: Bode plot generated from responses of neuron to planar tilt. Gain (relative to stimulus position; top) increased sharply as stimulus frequency was increased, and response phase (bottom) led stimulus position by ~90° at all frequencies. These response characteristics are similar to those of semicircular canal afferents.

View larger version (17K): [in this window] [in a new window]   Fig. 6.   Bode plot indicating mean gain (top) and phase (bottom) of responses to planar tilt of 12 neurons classified as receiving predominant otolith inputs and 7 neurons classified as receiving predominant inputs from semicircular canals. Not every neuron was tested at each frequency; however, all but 2 neurons with predominant otolith inputs and all but 1 neuron with predominant semicircular canal inputs were studied at 3 or more frequencies (typically including 0.1 and 0.5 Hz). Error bars indicate SE and are so small in some cases that they are not obvious. Mean response gain of neurons classified as receiving predominant otolith organ inputs increased <3-fold as stimulus frequency was increased from 0.05 to 0.5 Hz; response phase was near stimulus position at low frequencies but developed a progressive lag as stimulus frequency increased. In contrast, mean response gain of neurons with predominant semicircular canal inputs increased almost 10-fold as stimulus frequency was increased from 0.1 to 1 Hz, and mean response phase was advanced nearly 90° from stimulus position (i.e., near stimulus velocity) at all frequencies.

Locations of respiratory neurons with vestibular inputs. The locations of most respiratory neurons tested for vestibular inputs were reconstructed by reference to the placement of lesions near the recording sites. In some cases, however, the locations of lesions could not be determined, and thus not all of the recording sites were identified. However, the depth and laterality of all respiratory neurons studied suggest that they were in the VRG. The locations of E and I neurons that responded to tilt as well as those without vestibular inputs are indicated in Fig. 7. There was no obvious segregation of neurons with particular types of vestibular inputs in particular regions of the VRG.

View larger version (29K): [in this window] [in a new window]   Fig. 7.   Locations of I (top) and E (bottom) cells that were tested for responses to tilt and could be reconstructed from positions of lesions. Neuronal locations are plotted on transverse sections of medulla. Values to right of each section indicate relative distance (in mm) posterior to stereotaxic zero; level of obex was at ~P13.5. , Neurons whose activity was modulated by vestibular stimulation; , neurons whose activity was not affected by tilt. A, nucleus ambiguus; AP, area postrema; C, cuneate nucleus; DMV, dorsal motor nucleus of vagus; EC, external cuneate nucleus; G, gracile nucleus; IO, inferior olivary nucleus; IVN, inferior vestibular nucleus; LRN, lateral reticular nucleus; SNV, spinal trigeminal nucleus; ST, solitary tract; STV, spinal trigeminal tract; XII, hypoglossal nucleus.

Comparison of responses to tilt of spinal respiratory nerves and VRG neurons. Our previous studies showed that vestibular stimuli 10-15° in amplitude produce robust changes in the activity of nerves innervating abdominal muscles (16). To directly compare vestibular-elicited responses in VRG neurons and a spinal respiratory nerve, activity was recorded from an L₁ abdominal nerve during tilt in four animals. In all cases, 10° tilt produced modulation of abdominal nerve activity. In three animals, the direction of tilt producing maximal abdominal nerve responses was near nose-up pitch (96°, 104°, and 116°). In the fourth animal, the response vector orientation was near ipsilateral ear-down roll (2°). Responses recorded from this latter animal during roll tilt at a variety of frequencies are illustrated in Fig. 8. The response gain was relatively consistent across stimulus frequencies, and the response phase was near stimulus position at all frequencies. These characteristics of abdominal nerve responses to tilt are similar to those reported previously (16). For example, in this prior study, the L₁ abdominal nerve responses to vestibular stimulation had a vector orientation near nose-up pitch in about three-fourths of the animals and near ipsilateral ear-down roll in the other one-fourth. Furthermore, abdominal nerve responses to vestibular stimulation were previously reported to occur in phase with stimulus position at frequencies up to 1 Hz and to have gains that were flat across stimulus frequencies (16).

View larger version (32K): [in this window] [in a new window]   Fig. 8.   Responses recorded from an L1 abdominal nerve during vestibular stimulation delivered at an amplitude of 10°. A: responses to roll tilt at frequencies ranging from 0.05 to 0.5 Hz. Number of sweeps averaged in each trace: 15 at 0.05 Hz, 30 at 0.1 Hz, 51 at 0.2 Hz, and 113 at 0.5 Hz. B: Bode plot indicating gain (top) and phase (bottom) of responses relative to stimulus position. Gain was normalized (by dividing gain at each frequency by maximal response gain recorded). Response gain was flat across stimulus frequency, and response phase was consistently near stimulus position. These characteristics of vestibular-abdominal responses are similar to those reported previously (16) and are similar to those of otolith afferents.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

Previous experiments demonstrated that stimulation of vestibular otolith receptors produces changes in activity of respiratory nerves (16, 17). However, the present study showed that most respiratory neurons in the VRG, including E neurons that project to the spinal cord, do not respond to tilts up to 15° in amplitude. This finding suggests that neurons in addition to those in the VRG must be involved in relaying vestibular signals to spinal respiratory motoneurons, particularly to abdominal motoneurons, which have been shown to be powerfully affected by small-amplitude vestibular stimulation (16). Thus the present data support the conclusions of previous ablation studies, which demonstrated that chemical lesions of the VRG or transections of the axons of VRG E neurons do not abolish respiratory nerve responses to stimulation of the vestibular nerve (16, 20, 30).

Most of the I and E neurons in the VRG whose activity was modulated by tilt had different patterns of responses to vestibular stimulation than did respiratory nerves. The direction of vertical vestibular stimulation producing a maximal change in activity of VRG neurons was highly variable from cell to cell. In contrast, in most animals, nose-up pitch is the most effective direction of rotation for producing a change in activity in the phrenic, abdominal, and hypoglossal nerves (16, 17). Furthermore, although a few neurons may have been misclassified as receiving otolith organ, semicircular canal, or convergent otolith and canal inputs, it is also clear that the dynamics of responses of most VRG neurons to tilt differed from those of respiratory nerves. The responses of the hypoglossal, phrenic, and abdominal nerves to natural vestibular stimulation suggest that the predominant vestibular influences on these respiratory outflows come from otolith organs (16, 17). In contrast, over 40% of VRG neurons had response dynamics like semicircular canal afferents. Even those VRG neurons that apparently received otolith inputs had different responses to vestibular stimulation than did respiratory nerves. The response phases of VRG neurons classified as receiving predominant otolith inputs lagged stimulus position by an average of over 70° at frequencies 0.5 Hz. In contrast, vestibular responses recorded from respiratory nerves have response phases near stimulus position at all frequencies tested (up to 1 Hz) (16). It thus seems unlikely that even those VRG neurons that respond to tilts 15° in amplitude contribute appreciably (either directly or indirectly) to the production of vestibular-elicited responses in spinal respiratory nerves.

The present data thus raise three questions. 1) Why does electrical stimulation of the vestibular nerve affect the activity of a large number of VRG neurons, including those projecting to the spinal cord (12, 20), but moderate-amplitude natural vestibular stimuli do not alter the firing of these cells? 2) Which cells relay vestibular signals to spinal respiratory motoneurons? 3) What is the function of vestibular inputs to VRG neurons?

The answer to the first question may lie in the fact that electrical vestibular stimulation synchronously activates all vestibular afferents and thus provides a much stronger input to the central nervous system than was produced by the 7.5-15° tilts employed in the present study. It is thus possible that larger-amplitude vestibular stimuli would have activated more VRG neurons and that these neurons may contribute to vestibular-respiratory responses elicited by large changes in body position. It is also possible that the delivery of vestibular stimuli during only one phase of the respiratory cycle would have been more effective in producing responses in VRG neurons. Nonetheless, the present data clearly demonstrate that most spinally projecting VRG neurons are not affected by the same stimuli that elicit responses in spinal respiratory nerves, indicating that premotor neurons outside of the VRG must play an important role in producing the vestibular-respiratory reflex.

The answer to the second question is not available, since the locations of neurons outside the VRG that may contribute to vestibular-respiratory responses have not been fully identified. In general, although some data are available in the rat (4), little is known about neurons outside of the VRG and DRG that may provide inputs to respiratory motoneurons, particularly in emetic species. It is also feasible that neurons in the main respiratory groups which lack a spontaneous respiratory rhythm provide inputs to respiratory motoneurons. Further research will be required to determine which populations of brain stem neurons may relay vestibular signals to spinal respiratory motoneurons.

The role of vestibular inputs to the fraction of VRG neurons whose activity is modulated by moderate-amplitude tilts (third question raised above) also awaits determination. Because both the spatial and temporal characteristics of the responses of VRG neurons to vestibular stimulation differ substantially from those of the hypoglossal, abdominal, and phrenic nerves, it seems unlikely that cells in the VRG are involved in producing vestibular-respiratory responses in these nerves. However, the responses of many respiratory nerves to natural vestibular stimulation, including those innervating laryngeal, pharyngeal, and intercostal muscles, are yet to be determined. Previous studies employing electrical stimulation of vestibular afferents (21, 28) have shown that the laryngeal, pharyngeal, and intercostal muscles are influenced by the vestibular system. It is possible that VRG neurons have response properties which match those of these additional muscles and that cells in the VRG relay vestibular signals to pharyngeal, laryngeal, or intercostal motoneurons; this prospect remains to be explored.

Perspectives